Optical Coherence Tomography May Be Used to Predict Visual Acuity in Patients with Macular Edema

Abstract

Purpose.: To determine whether the volume of retinal tissue passing between the inner and outer retina in macular edema could be used as an indicator of visual acuity.

Methods.: Diabetic and uveitic patients with cystoid macular edema (81 subjects, 129 eyes) were recruited. Best corrected logMAR visual acuity and spectral optical coherence tomography (OCT/SLO; OTI, Toronto, ONT, Canada) were performed in all patients. Coronal OCT scans obtained from a cross section of the retina between the plexiform layers were analyzed with a grid of five concentric radii (500, 1000, 1500, 2000, and 2500 μm centered on the fovea). The images were analyzed to determine the amount of retinal tissue present within each ring. A linear regression model was developed to determine the relationship between tissue integrity and logMAR visual acuity.

Results.: A linear relationship between tissue integrity and VA was demonstrated. The volume of retinal tissue between the plexiform layers in rings 1 and 2 (up to 1000 μm from the foveal center) predicted 80% of visual acuity. By contrast, central macular thickness within the central 1000 μm predicted only 14% of visual acuity.

Conclusions.: This study showed that the cross-sectional area of retinal tissue between the plexiform layers in cystoid macular edema, as imaged by OCT, is the best indicator of visual function at baseline. Further prospective treatment trials are needed to investigate this parameter as a predictor of visual outcome after intervention.

Macular edema results from abnormal accumulation of fluid in the central retina and indicates compromised function in one or both of the blood–retinal barriers. It is a common sequel of many ocular conditions and the main cause of visual loss in diabetic retinopathy. ^{1 –5}  

Any abnormal pooling of extracellular fluid may result in displacement of the spatial relationships between retinal neuronal components. Small amounts of fluid may lead to an increase in overall retinal thickness, whereas larger amounts may give rise to cell-free spaces as seen in cystoid macular edema (CME). ⁵  

There is an extensive body of literature detailing the association between central macular thickness (CMT) and visual acuity (VA), ^{6 –9} with further studies claiming the beneficial effects of treatments designed to reduce CME. ^{10 –15}  

Observations from histology and optical coherence tomography (OCT) of macular edema give a false impression of multiple cysts delineated by tissue structures in the z-plane of the retina. However, scanning electron microscopy shows that more commonly a single cystic space is present within which a number of structures extend from the inner to the outer retina. Such structures consist of columns of Müller's fibers together with the axonal elements of bipolar cells passing between the two plexiform layers. ^{16 –18} Empiric studies have demonstrated that the two plexiform layers together with the outer limiting membrane form a physical resistance barrier to fluid movements. ¹⁹ Thus, extracellular fluid may be contained within layers defined by these resistance barriers. In diabetic retinopathy, cystic spaces may occur either between the inner and the outer plexiform layers or between the outer limiting membrane and the outer plexiform layer. In the former location, there is a potential to displace bipolar cells leading to cell loss or compromised function, while in the latter, only photoreceptor cells are at risk. ^20,21  

Given the fundamental role of bipolar cells in being the sole communication pathway between photoreceptors and ganglion cells, any loss of connectivity between these cells will compromise visual function. It follows that the more the retinal thickness increases, the more such axons will be stretched and as a consequence, some will break. This mechanism is probably the one that underlies the apparent relationship between increasing retinal thickness and decreasing VA. By contrast, those bipolar cells whose axons are closely adjacent to Müller's fibers will have a greater chance of surviving displacement because of the greater physical strength and support provided by the adjacent Müller's fibers. ^16–17  

Theoretically, a useful indicator of the VA and potential visual outcome in eyes with macular edema would be to analyze the residual volume of tissue passing between the two plexiform layers, as only such areas would allow passage of bipolar axons between photoreceptors and ganglion cells. The impact of photoreceptor–ganglion cell connectivity on VA would further depend on the location of surviving axons within the central visual field. ²² Thus, an optimal measurement of potential function would be an evaluation of the number of vertical elements passing between the plexiform layers, their diameter, and their eccentricity from the fovea. 

This study had two objectives: to assess new hardware for imaging retinal glial tissue in cases of CME and to investigate whether the amount of glia and by association residual bipolar cells could be used as indicator of visual function in patients with CME. 

Methods

Experimental Design

Patients with macular edema were prospectively recruited from both diabetic and uveitic outpatient clinics over a period of 9 months. The study involved a baseline assessment of visual function, ophthalmoscopy, and OCT imaging at a single time point. The research conformed to the tenets of the Declaration of Helsinki, with informed consent being obtained from the subjects subsequent to explanation of the nature of the study. The protocol of the study was approved by the local ethics committee (protocol number, 06/Q0702/175). 

Patient information was anonymized at the time of patient recruitment, to allow independent data analysis. 

Inclusion criteria for the study were a clinical diagnosis of CME, confirmed either by OCT alone or by OCT and fundus fluorescein angiography (FFA) at the time of enrollment. For each patient, either one or both eyes were included in the study. 

Patients with coexisting ocular diseases were excluded. Exclusion criteria included the presence of media opacity affecting the quality of the OCT scan and angiographic or clinical evidence of ischemic maculopathy. Patients were also excluded if they had undergone major ophthalmic surgery including cataract procedure or had received laser treatment in the previous 4 months. None of our patients had received anti-VEGF treatment. 

Subjects and Clinical Procedures

Each patient underwent a complete anterior segment examination by slit lamp biomicroscopy and best corrected VA assessment using a logMAR chart at 3 m distance. All eyes were then dilated with phenylephrine 2.5% and tropicamide 1% and examined by indirect funduscopy with a 78-D lens. In the diabetic patients, fluorescein angiography was required as one of the inclusion criteria of the study, to assess retinal circulation and allow exclusion of patients with subclinical foveal ischemia. In cases in which a fluorescein angiogram could not be performed, recruitment into the study was allowed when clinical findings excluded ischemia. 

Optical Coherence Tomography

OCT was performed with a spectral domain OCT/SLO system (Spectrum-OTI Spectral OCTSLO model E; Spectrum-Ophthalmic Technologies Inc., Toronto, ONT, Canada). This device is an optical imaging system, combining a confocal scanning ophthalmoscope and OCT. Both the confocal fundus SLO image and the OCT image are generated through the same optics and displayed simultaneously on the computer screen with pixel to pixel correspondence. The dual-imaging system enabled simultaneous images to be presented thus the point of fixation could be accurately cross-correlated with the OCT image and any loss of fixation was immediately detected and the corresponding OCT discarded. The system uses light generated from an infrared broadband super luminescent diode (SLD) with a wavelength between 790 and 950 nm. Cross-sectional images of the retina along the x–y plane (B-scan), such as single line, radial, and raster scans, could be obtained, as well as coronal images within the z plane (C-scan). 

In preliminary studies, scans were obtained in three different modes of operation. First, a series of 24 radial scans over 360° was automatically initiated, intersecting at the center of the patient's fixation. Second, a single scan mode was selected whose orientation and location within the fundus was determined by the operator. Third, the system was used to generate a raster scan of the macula from the superior to the inferior arcade with 64 scans, again centered by the patient's fixation. Given that radial scans and to a lesser extent raster scans always have a sampling error, we always included C-scans in our analysis. 

Three dimensional views of the macula were obtained by selecting the topography mode where images were viewed as surface maps and these were extracted manually by slicing the 3-D picture using the device-based image analysis software. 

Macular thickness measurements were derived from the topography scan by selecting the Macular Thickness Map function and recording the central subfield mean thickness (CSMT). The location of the central subfield mean thickness was identical with that used in the ETDRS grid of macular zones. ²³  

Coronal scans (C-scans) were fundamental to the present study and were obtained by selecting the midpoint between the ganglion cell layer and the innermost aspect of the outer plexiform layer, in most cases to mid depth of the cysts. In practice this was obtained by adjusting the section plane to an appropriate level parallel with the retinal surface in the B scan displayed on the x-axis of the coronal image. 

Image Analysis

To investigate the purpose of this study, three pieces of information were required from each patient: the number of columns of tissue present, their cross-sectional area at the narrowest point, and the eccentricity from the foveal center. 

An image analysis system was created to extract each of these datasets. Data were collected from a series of concentric rings of 500, 1000, 1500, 2000, and 2500 μm radii. Thus, within each ring, the area of surviving tissue as opposed to cystic space was extracted by first compressing the gray scale such that tissue appeared white and edema black. ²⁴ The second part of the program counted the number of white pixels present within each ring, thus giving a measure proportionate to the potential number of connections passing between the two plexiform layers. The number of pixels of spared tissue within each annulus was converted to an area in square millimeters by scaling the ratio of the number of pixels of spared tissue to the total number of pixels in the annulus by the area in square millimeters of the annulus. 

Outcome Measures

This study had three primary outcomes: best corrected logMAR VA; retinal tissue integrity evaluated as number of pixels corresponding to the tissue component between cystic spaces and observable at increasing eccentricities from the fovea in segmented images of OCT/SLO coronal scans; and CMT measurement obtained from the OCT/SLO retinal thickness map. 

Statistics

All data organization and manipulation was performed in a spread sheet program (Office Excel 2003; Microsoft, Redmond, WA) and statistical analysis was performed in commercial software (SPSS 16.0 for Windows; SPSS Inc, Chicago, IL, and Minitab ver. 13.30; Minitab Inc, State College, PA). After tests for normality (Kolmogorov-Smirnov test; P < 0.05 indicated a significant difference from normal), correlation coefficients were calculated to evaluate the association between VA and the other six outcome measures. 

A linear regression model was developed to assess whether the amount of glia could be used to predict VA. The data set of 129 eyes were randomized and split into two data sets, one of 100 eyes and the other of 29 eyes. A stepwise linear regression was performed on the data set of 100 eyes with logMAR VA as the dependent variable and the other six outcome measures as predictors. The criterion for entry into the model was P = 0.05 and 0.10 for removal. Stepwise linear regression is an extension of simple linear regression where the dependent variable is predicted by a linear equation involving one outcome or independent variable and a constant. In stepwise linear regression, multiple variables can be linearly combined in the model. They are entered automatically by the statistics software provided they make a statistically significant improvement in the model. Stepwise linear regression has been used in a large number of areas. ²⁵  

The remaining 29 eyes were used to test the model by assessing the agreement between the predicted and measured VA using the Bland-Altman method. ²⁶  

Results

For classification, Figure 1 demonstrates the limited spatial information in a two-dimensional image, as provided by both conventional histology and OCT. Both images show a series of compartments divided by vertical structures running between the two plexiform layers of the retina. By contrast, the three-dimensional information in Figure 2 demonstrates a single space of considerable volume within which there is a three-dimensional distribution of predominantly Müller fiber columns passing between the two plexiform layers. The high magnification detail in Figure 2 further demonstrates the association of neuronal elements with such structures. Figure 3 illustrates two cases of diabetic macular edema with similar CMT measurements and different amounts of residual tissue between the plexiform layers on B- and C-scans. LogMAR VA in the two individuals in Figures 3A and 3B was +0.3 and +1.0. Figure 4 illustrates a macular thickness map (Fig. 4A) and an example of coronal scan with superimposed concentric rings of 500, 1000, 1500, 2000, and 2500 μm in radius (Fig. 4B). 

Figure 1.

 

Light microscopy and OCT (OCT/SLO; OTI, Toronto, ONT, Canada) images of human retina affected by CME. Examples of early (top) and late (bottom) CME are represented. Intraretinal fluid appeared to be contained in cystic spaces separated by walls.

Figure 2.

 

Scanning electron microscopy of CME. Columns of tissue are standing up in a continuous space of fluid pooling. Retinal elements along the z-plane are represented by bipolar axons and Müller fibers (author JM's collection). Right: high-magnification detail.

Figure 3.

 

(A) OCT B- and C-scan of a patient with diabetic macular edema, +0.3 logMAR VA, and 538 μm CMT. This case illustrates the high amount of residual tissue between the plexiform layers. (B) OCT B- and C-scans of a patient with diabetic macular edema, +1.0 logMAR VA, and 522 μm CMT. This case is an example of a very low amount of residual tissue between the plexiform layers.

Figure 4.

 

(A) Macular thickness map of a patient with macular edema representing subfield mean thicknesses, as from an ETDRS study. (B) Gray scale coronal OCT scan with superimposed grid dividing the macula into five areas of increasing eccentricity (radii: 500, 1000, 1500, 2000, and 2500 μm).

Patients

A total of 81 participants were enrolled: 36 men and 45 women. The average age was 63 years (range, 26–87 years; Table 1). 

Table 1.

 

Patient Characteristics

Patients (n = 81)	Diagnosis	Unilateral (n = 33)	Bilateral (n = 48)
60	CME	14 cases	46 cases
12	Uveitis	10 cases	2 cases
6	RVO	6 cases	0 cases
2	Irvine-Gass	2 cases	0 cases
1	Tractional	1 case	0 cases

 RVO, retinal vein occlusion.

Most patients (73%, 59 subjects) underwent fluorescein angiography, but, in the remaining group, an angiographic study could not be performed (27%, 22 subjects) because of previously documented adverse reactions to the dye (9 subjects), refusal to participate in the investigation (8 subjects), or technical difficulty in obtaining satisfactory venous access (7 subjects). Typical patient contact time was 40 minutes, of which only 5 minutes was needed for OCT imaging. 

Relationship between Tissue Integrity and Visual Function

The scatterplots in Figures 5B and 5C show a linear relationship between the amount of spared tissue within rings 1 and 2 and logMAR VA. The correlation falls progressively for rings 3, 4, and 5 (Figs. 5D–F). Not all variables were normally distributed (P = 0.000–0.200); therefore, Spearman rank correlation coefficients were calculated. Their values, together with the associated statistical significance and the R ² values are given in Table 2. Figure 6 illustrates the R ² values reported as percentages of variation in VA explained by the outcome measures (CMT or the tissue sparing within each of the five rings). 

Figure 5.

 

Scatterplots showing the relationship between (A) CMT versus logMAR VA (r _s = 0.407*); (B) tissue integrity within circle 1 versus logMAR VA (r _s = −0.832*); (C) tissue integrity within circle 2 versus logMAR VA (r _s = −0.841*); (D) tissue integrity within circle 3 versus logMAR VA (r _s = −0.624*); (E) tissue integrity within circle 4 versus logMAR VA (r _s = −0.277*); and (F) tissue integrity within circle 5 versus logMAR VA (r _s = −0.134*). *Significant correlation at the 0.05 level (two-tailed).

Table 2.

 

Regression Values and Spearman's Correlation Coefficients Describing the Relationship between the Outcome Measures and LogMAR Visual Acuity for all 129 Eyes in the Study

Variable	r s	P (Two-Tailed)	R 2 (%)
CMT	+0.407	<0.001	16.6
Tissue spared in ring 1 (500 μm)	−0.832	<0.001	69.2
Tissue spared in ring 2 (1000 μm)	−0.841	<0.001	70.7
Tissue spared in ring 3 (1500 μm)	−0.624	<0.001	38.9
Tissue spared in ring 4 (2000 μm)	−0.277	0.001	7.7
Tissue spared in ring 5 (2500 μm)	−0.134	0.130	1.8

Figure 6.

 

Variation in R ² values representing the association between VA and retinal spared tissue at increasing eccentricity as well as VA and CMT.

Relationship between Macular Thickness and Visual Function

The relationship between CMT and VA is shown in the scatterplot in Figure 5A. The correlation between CMT and logMAR VA was moderate (r _s = 0.407). The coefficient of determination (R ²) demonstrates that CMT only explains 16.6% of the change in VA, according to a linear model (Table 2). 

Linear Regression Model

The linear regression model after stepwise linear regression was given by  

 

where CMT is the central macular thickness (millimeters), and T ₁ and T ₂ are the areas of tissue sparing in square millimeters in rings 1 and 2, respectively. This model had an R ² value of 80.7%, indicating that equation 1 accounted for more than 80% of the variation in logMAR VA. It is noteworthy that the most predictive variable was T ₂, and this alone could predict 74.4% of the variation in logMAR VA in a linear model. The significance of this result will be commented on further in the Discussion section. 

Figure 7 shows the results of testing the model and is a scatterplot of the measured logMAR VA plotted against the estimated logMAR VA, according to equation 1. A line of equality is shown along which all data points would be expected to lie in the presence of perfect agreement. Clearly, this was not the case, as expected from most clinical measures. 

Figure 7.

 

Scatterplot, with line of equality, for measured versus predicted logMAR VA using the linear regression model developed in this study.

Figure 8 shows the Bland-Altman mean difference plot for our data. It demonstrates a bias of −0.02 logMAR units and limits of agreement of ±0.23 logMAR units. 

Figure 8.

 

Bland-Altman mean difference plot demonstrating agreement between measured and predicted logMAR VA.

Discussion

This study demonstrated that there is a strong correlation between VA in patients with CME and the volume of tissue passing between the two plexiform layers in the central retina as determined by OCT. It is the first time that a predictive measure of visual performance has been derived from imaging of macular edema. 

The results demonstrate that good VA occurred only in those patients with an adequate volume of tissue running between the inner and the outer plexiform layers in the central 1000 to 2000 μm of retina (Figs. 5, 6). Given that foveal cones have inner connecting fibers that may be up to 500 μm in length, ²² foveal cones may connect to bipolar cells displaced 500 μm radially from the inner and outer segments. Thus, this lateral displacement of connections between foveal cones and ganglion cells explains the dependency of VA on the tissue integrity in both rings 1 and 2. 

Although there was still reasonable correlation within ring 3, presumably due to signals derived from photoreceptors at the extreme edges of the fovea, correlation was lost within rings 4 and 5. In these locations, although large amounts of tissue volume may be spared, the connectivity is predominantly with extrafoveal photoreceptors. 

These findings are in keeping with both an anatomic study looking at displacement of retinal ganglion cells subserving cones in the human fovea ²² and a pathologic study measuring laser damage to foveal photoreceptor cells. ²⁷ Sjöstrand et al. measured the radial offset produced by cone fibers within the layer of Henle and demonstrated that, at the foveal border (0.5–0.8 mm or 1.8–2.9° eccentricity), the mean offset due to the fibers of Henle and the mean total lateral displacement were at a maximum of 0.32 ± 0.03 and 0.37 ± 0.03 mm, respectively, thereafter steeply decreasing outside the foveal border to an eccentricity of 2 to 2.5 mm. This anatomic finding confirms that structural damage involving the neural retina up to 1 mm from the foveal center may have implications for loss of information generated within the fovea. 

To estimate the number of axonal elements necessary to maintain the potential for visual function, more information is needed with regard to the spatial arrangement of bipolar and Müller fibers in the residual columns and their relative ratio. From electron microscopy studies, it is known that bipolar axons are surrounded by Müller fibers in the human retina. ²⁸ Given that the approximate diameter of bipolar axons is 0.5 μm and the diameter of Müller fibers is between 5 and 10 μm, we can estimate that, in normal conditions for each Müller fiber, there are approximately 34 to 68 adjacent bipolar neurons around their circumference. 

By contrast, in pathologic conditions such as macular edema, accumulation of intracellular fluid may reduce the total number of Müller fibers but may also result in an increase in the diameter of those remaining up to 15 to 20 μm. This would allow an increase in the potential number of surrounding bipolar axons to approximately 97 to 128 (Fig. 2). 

Furthermore, from the linear regression model, it appears that a minimum of 50% of preserved retinal tissue within ring 1 is necessary to maintain a VA of 0.4 logMAR or better (Fig. 5B), whereas at least 70% of the retinal tissue within ring 2 is necessary for a level of VA of 0.4 logMAR or better (Fig. 5C). 

Even though the total number of bipolar axons traversing the space between the plexiform layers may be significantly reduced, both horizontal and amacrine cells will contribute to image processing and VA by integrating signals over a number of photoreceptor cells and ganglion cells, respectively. 

The major source of error in the present study is the potential loss of information implicit in compression of gray scale during the image analysis. At extremes, potential connectivity may be lost if too few pixels are present in a given element within the primary image to be resolved as tissue. By contrast, some discontinuous elements which do not really traverse the interplexiform space may be interpreted as intact elements. Further work is therefore needed to optimize the image analysis process and to give further maximum and minimum correlates for axonal elements potentially associated with the glial tissue. 

In the present study, no attempt was made to determine degradation of visual information resulting from fluid accumulation, either beneath the retinal pigment epithelium or beneath the interphotoreceptor matrix. Clearly, the potential VA of any given patient relates to both the quality of the image presented to the photoreceptor cells and the ability to transmit such resolved information from the photoreceptors to the ganglion cells. 

The former may well be degraded by the presence of fluid in the outer retina, whereas the latter will be less dependent on fluid distribution in the inner retina and more dependent on the number of viable bipolar axons. 

Recovery or improvement of VA may be dependent on changes induced in both these components. First, some improvement may occur resulting from removal of fluid from the outer retinal layers as a consequence of changing fluid and ionic environment in the interphotoreceptor matrix thus potentially enhancing transduction. Second, whereas no current therapeutic regimen will replace lost neurons, drainage of the cystoid fluid will ease the tension on the axons passing through the cysts and at the same time change the ionic environment around the axons and both of these may enhance function. Given the role of the Müller fibers in transretinal transport of ions, ²⁹ changes to their environment maybe of particular importance. 

At present, the measurement of retinal thickness with OCT represents the accepted standard despite its incredibly poor correlation with visual function (R ² values ranging from 0.08 to 0.54) ⁹ compared with the results presented in the current paper. 

The relationship between patterns of macular edema and VA has also shown poor consistency across different studies. ^{30 –32} The apparent correlation between the increase in retinal thickness and the decrease in VA may be explained by the results of the present paper whereby increase in thickness will be associated with increase in loss in viable axons. The more direct approach to assessing neuronal survival in the present study would also explain why the correlation values are so much better. 

The neuroretina is a complex matrix within which the only connection between photoreceptors and ganglion cells are bipolar cells. Like any biological material the system has a degree of elasticity and therefore as a consequence fluid pooling in the interplexiform layer will result in an overall increase in retinal thickness. Within elastic limits, bipolar continuity will be maintained. However, if sufficient swelling occurs that exceeds the elastic limits, then bipolar axons may snap and the transmission pathway may be lost. This concept would explain the apparent correlation between thickness and VA. In the present study, by measuring tissue available for connectivity in relation to eccentricity, the bipolar survival concept is refined and the potential value of this measurement is demonstrated in Figure 6. 

The present study validates retinal tissue integrity as a measure of preserved axonal connections and indicator of visual function. The strength of the relationship between preserved tissue and visual function, as expected, decreases at increasing eccentricities from the center of the fovea. 

Finally, it should be remembered that the OCT gives only spatial information on anatomic structures and while the correlation derived in the present paper has given an excellent correlation with potential VA, this study has not and cannot demonstrate physiological changes that may also correlate to visual loss in these patients. 

It will be of interest to apply the current analysis to future trials of regimens designed to modulate edema. The ability to determine the potential visual outcome for patients before the commencement of any treatment trial will be highly beneficial, in that it will allow exclusion of those individuals who could not in anyway benefit from intervention. A longitudinal treatment trial is necessary to validate the assessment of retinal structural integrity at baseline and to relate this parameter to the final visual outcome. 

Footnotes

 Supported by a grant from The Royal College of Surgeons, London, United Kingdom.

Footnotes

 Disclosure: L. Pelosini, None; C.C. Hull, None;, J.F. Boyce, None; D. McHugh, None; M.R. Stanford, None; J. Marshall, Ophthalmic Technologies, Inc. (C)